> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cometly.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Data Warehouse Exports

> Export your raw Cometly data as Parquet files to your own S3 bucket for loading into a data warehouse or lake.

## Overview

Data Warehouse Exports write your raw Cometly data as **Parquet files** to an **S3 bucket you own**. Point your warehouse or lake tooling — Snowflake, BigQuery, Databricks, Athena, DuckDB, and others — at the bucket to load the data on your own schedule.

You keep the raw, unaggregated data and model it however you like on your side.

<Note>
  This is different from the **API exports** in the API Reference
  ([Create Event Export](/api-reference/endpoint/create-event-export),
  [Create Contact Export](/api-reference/endpoint/create-contact-export),
  [Create Company Export](/api-reference/endpoint/create-company-export)). Those return gzipped NDJSON
  for ad-hoc download through the API. Data Warehouse Exports deliver columnar Parquet files to your
  own object storage for bulk, recurring loads. See
  [Exports vs Pagination](/concepts/exports-vs-pagination) for the API-side options.
</Note>

## How it works

<Steps>
  <Step title="Add a destination">
    Configure the S3 bucket and credentials Cometly writes to. See [Set up an S3 destination](/data-warehouse/s3-destination).
  </Step>

  <Step title="Create an export">
    Choose a destination, a dataset, a cadence, and a window. See [Create an export](/data-warehouse/create-export).
  </Step>

  <Step title="Load the files">
    Cometly writes Parquet files under your chosen path. Point your warehouse or lake at the bucket to load them.
  </Step>
</Steps>

## Datasets

Each export unloads exactly **one** dataset. To export several datasets, create several exports — each with its own cadence, window, destination, and columns.

| Dataset                                                                   | Description                                                                                                              |
| ------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ |
| [Raw touchpoints](/data-warehouse/datasets/raw-touchpoints)               | Every tracked touchpoint as recorded, over the window.                                                                   |
| [Raw conversions](/data-warehouse/datasets/raw-conversions)               | Every conversion as recorded — the unattributed counterpart to touchpoints.                                              |
| [Raw hits](/data-warehouse/datasets/raw-hits)                             | Every tracked pixel hit as recorded, unfiltered across all event types.                                                  |
| [Attributed conversions](/data-warehouse/datasets/attributed-conversions) | Conversions joined to their attributed touchpoints, at the attributed-touchpoint grain.                                  |
| [Contacts](/data-warehouse/datasets/contacts)                             | Every contact created in the window, with its identity and custom fields.                                                |
| [Companies](/data-warehouse/datasets/companies)                           | Every company created in the window.                                                                                     |
| [Ad analytics](/data-warehouse/datasets/ad-analytics)                     | Daily ad-performance rows per channel (Facebook, Google, LinkedIn, TikTok, Microsoft, Reddit, Taboola, AdRoll, Whop, X). |

## Cadence and windows

An export runs on a **cadence** and covers a **window** of data. The two are independent: cadence is *how often* a run fires; the window is *how much* each run pulls.

| Cadence      | Runs                              | Window                                                                       |
| ------------ | --------------------------------- | ---------------------------------------------------------------------------- |
| **One-time** | Once, then stops                  | An absolute start/end date range you pick.                                   |
| **Daily**    | Once a day, at an hour you choose | A trailing window (e.g. last 7 days), capped at **90 days**.                 |
| **Hourly**   | Every hour                        | A trailing window, capped at **14 days** (7 days for ad-analytics datasets). |

A **trailing window** is anchored to "now" at run time — "last 7 days" means the 7 days ending at the moment the run fires, not a fixed calendar range. Every recurring run re-exports the **full** window (not just what changed), so runs are self-healing and you deduplicate on a stable key on your side.

<Note>
  A run only re-exports rows that are still **inside** the window. To refresh data older than the
  window (for example, after a backfill or a late restatement), create a **one-time** export over the
  date range you want. The common pattern is a one-time backfill plus a recurring incremental export.
</Note>

## Column conventions

These conventions hold across every dataset, so the same column name means the same thing in every export.

* **Timestamps are UTC.** Every `*_utc` column (and `created_at` / `updated_at`) is a UTC timestamp. The one exception is the ad-analytics **metric date**, which is a local calendar date — see that dataset's page.
* **Locked vs optional columns.** Some columns are **locked** (always exported — the ones that keep the data coherent, such as row identity and join keys); the rest are **optional** and you pick them at creation. Each dataset page marks which is which.
* **Cometly ID vs platform ID.** Ad entities (account, campaign, ad set/ad group, ad) are carried as **IDs only — never names**. Each level exposes two columns:

  * `cometly_<level>_id` — Cometly's own internal ID for the entity. Stable, and the reliable key for joining one export to another (e.g. touchpoints to ad analytics).
  * `platform_<level>_id` — the advertising platform's own ID for the same entity (the Facebook/Google/etc. ID). The key you join to your platform-side data; may be null.

  Resolve campaign/ad **names** on your side from these IDs.
* **JSON-array columns.** A few columns (the contacts identity-graph columns) are exported as JSON-encoded arrays. They're optional and default to off.

## Where to manage exports

In the Cometly dashboard, go to **Integrations → Data Exports**. Destinations and exports are both managed there. This feature is available to space administrators.